Analyzing Gait With Minimal Body-Worn Sensing: Combining UWB and IMU

Typical applications of qualitative gait assessment in the daily life of individuals with gait disorders require multiple on-body sensors to determine spatial-temporal outcomes such as step length, width, and foot inclination/progression angles. These outcomes reflect the stability of their gait and can be used to provide the user with feedback on how to improve it.

However, requiring many body-worn sensors is expensive and impractical for the end user, especially in a real-world setting. Applications with a minimal set of on-body inertial measurement units (IMUs) have been developed, but they suffer from drift in relative foot positions or assume a symmetrical gait.

This limits the implementation of daily-life gait monitoring to specific individuals and short walking bouts. To address this gap, this research presents a method that combines data from sensors placed on both feet, an IMU with an ultra-wideband (UWB) receiver on one foot, and an IMU with a UWB sender on the other.

UWB data provides the relative distance between sender and receiver as well as the relative angle in the plane of the antennas (angle of arrival). This allows improved tracking of foot position and orientation by stabilizing the integration of accelerations (IMU data) with direct measurements of relative position/orientation (UWB data).

The researchers use an Extended Kalman Filter (EKF) approach that combines information from the sensors (IMUs and UWB), weighting their reliability, and estimates the position and orientation of the feet based on the weighted sensor data. To be more precise, integrating the acceleration data accumulates errors over time (drift), while the relative position measurements (from UWB data) help to stabilize that drift and reduce the positional error.

The approach was compared to using only IMUs on the feet by four healthy participants. It showed significantly improved spatial-temporal outcomes: step length error was 0.037 (relative to body height) compared to 0.198; step width error was 0.049 (relative to body height) compared to 0.456; and foot progression angle error remained similar to when only IMU sensors were used, at 3.4° compared to 3.3°. It was shown that the angle of arrival had minimal impact on accuracy due to inaccurate measurements. It is therefore expected to improve with software and hardware improvements of the UWB module.

These results are comparable to related work in the field that combines ultrasonic sensors with IMUs. However, ultrasound measurements are heavily influenced by occlusions, through-tissue transmission, temperature, and air pressure, which complicates their use in daily life. A different study that showed similar accuracy was only applicable to straight-line walking. The proposed method is therefore less challenging to apply to individuals with gait disorders.

This work used hardware that exhibited noticeable errors when the feet were close together. Future work should therefore focus on integrating improved UWB hardware (including additional antennas for 3D angle-of-arrival measurements) directly into IMU hardware as a single module. This approach is a promising development that could enable daily-life monitoring of individuals with gait disorders in ambulatory settings.